AUT - Centre for Neurocomputation and Neuroinformatics
Centre for Neurocomputation and Neuroinformatics
The Centre develops brain-like computational methods and systems for learning and knowledge discovery. The methods are applicable in the areas of neuroinformatics and neuroscience, as well as for solving complex problems of Artificial Intelligence, such as multimodal information processing.
People
- Prof. Nikola Kasabov (Centre Director) email: nkasabov@aut.ac.nz
- Prof. Valery Feigin (Associate Director), email: vfeigin@aut.ac.nz
- Dr. Lubica Benuskova
- Dr. Qun Song
- Mr. Simei Gomes Wysoski
- Prof. Ikuko Nishikawa
- Mr. Vishal Jain
- Mr Paulo Gottroy
External Collaborators
- Prof. M. Arbib, University of Southern California, US
- Dr. M. Paulin, University of Otago
- Prof. Takeshi Yamakawa, KIT, Japan
- Dr. Cees van Leeuwen, RIKEN Brain Science Institute, Japan
- Prof. Jagath Rajapakse, Nanyang University of Technology, Singapore
Current research topics include:
-
Modelling learning and memory mechanisms in the brain.
-
Spiking neural networks: methods, systems, implementations, applications.
-
Brain data analysis (EEG, fMRI).
-
Computational neuro-genetic modelling (CNGM): integrating gene and brain data to model brain functions (Fig. 1).
-
The world-first Brain-Gene–Simulation Ontology (BGO): A repository of conceptual and factual data, information, knowledge and models related to brain structures, functions, diseases and the genes/proteins involved in them. This is used for teaching and research.
Figure 1. The world-first dynamic computational neuro-genetic model (CNGM) developed at KEDRI consists of gene regulatory network to model gene interactions within a spiking neural network model. The model is optimised so, that the model behaviour matches real brain data.
Software System for Neuro Genetic Simulation
Project Team
- Mr. Simei Gomes Wysoski
- Dr. Lubica Benuskova
- Prof. N.Kasabov
A novel computational approach to neural network modelling that integrates dynamic gene networks with a neural network model was introduced in. In the model, interaction of genes regulates the activity of neurons that consequently affects the dynamics of the whole neural network. It has been shown that tuning the interaction between genes and the initial gene/protein expression levels, different states of the neural network operation can be achieved. The main goal of neurogenetic models is to simulate the activity of certain parts of brain using biologically plausible neural networks and the link to the genetic level, in an attempt to enable discoveries of yet unknown dynamic relationship between genes and states of the brain activity.
In this project, we present the first simulator for neurogenetic models. The biologically plausible neural network is implemented using spiking neural network (SNN). The behaviour of the SNN is evaluated by means of the spectral analysis of field potential and neural activity levels (spiking rate and neuronal synchronisations). The simulation of gene interactions is done through linear and non-linear gene regulatory network (GRN) models as proposed in or using two coupled delayed differential equations where delays between mRNA levels and proteins can be taken into account.
The simulator is implemented in C++ language. Preliminary tests have been performed using neuronal parameters extracted from the literature. Values for GRN were randomly generated, since most of the gene interactions related to brain activity still unclear in the literature. Using a simple algorithm of optimization (Genetic Algorithms) we were able to find sets of parameters that lead to a predefined behaviour.
Despite promising preliminary results, there are some issues that must be addressed to perform more realistic simulations, to include, a) Lack of information at the genetic level. Even with the recent genetic discoveries, GRN still in an initial stage of development. Further advances in this area are decisive to successfully perform simulations of neurogenetic models; b) Time Scale: SNN works in a time scale of milliseconds while GRN and the gene expression levels have much slower dynamics (minutes to hours). Since the simulation of SNN is very time consuming even for an ensemble of few hundreds of neurons, it still very difficult to perform the simulation for long periods of neuronal activity.
Brain-Gene Ontology (Homepage)
Project Team
“BGO” system is a teaching and research tool in neuroscience area that includes various concepts, facts, data, graphs, videos, animations, and other information forms, related to brain functions, brain diseases, their genetic basis and the relationship between all of them. Relationships are represented graphically that enables visualisation and creation of new relationships. Behind this system is Computational Neurogenetic Modelling (CNGM), a novel research attempt to simulate brain functions or a brain disease manifestation and ontology maps (linked to diseases) which provides a conceptual framework to store and access factual knowledge.
Please visit Brain-Gene Ontology's homepage for more details